GAO wants CMS to check whether hospitals' price transparency data are actually usable

The Biden administration could stand to take a firmer hand on hospital price transparency, especially when it is unclear whether the price data being published are even accurate, the Government Accountability Office (GAO) wrote in a Wednesday report.

The Centers for Medicare & Medicaid Services (CMS) has required hospitals to post the prices for numerous services annually and this past summer raised the bar by ensuring hospitals were doing so using a standardized file format.

Numerous reports from stakeholders criticized hospitals’ compliance along the way, with hospitals themselves often saying that the requirements were burdensome and often too vague.

On instruction from Congress, the GAO conducted a review of the requirements, the CMS’ enforcement and whether the agency’s policy was successfully serving patients, payers and researchers.

The GAO interviewed 16 stakeholder groups—representing those three groups—who described difficulties making effective comparisons and compiling the data for large-scale use. These hurdles were tied to inconsistent file formats, pricing complexities that came across poorly in the machine-readable format and what they perceived to be incomplete and inaccurate data sets.

“While the use of hospital price transparency data has been limited so far, many stakeholders we interviewed noted that they expect use to increase over time if the data usability challenges are overcome or addressed,” the GAO wrote in the report. “Further, some stakeholders also noted that it will take health plans and employers time to figure out how to effectively use the pricing data as part of their price negotiations and their efforts to develop networks of health care providers.”

When reviewing the CMS’ enforcement efforts, the office found that, from 2021 through 2023, the CMS had initiated 1,287 enforcement actions, about two-thirds of which came in the final year. The enforcement actions most often cited deficiencies related to missing data (43% of actions), no machine-readable file (34%) and noncompliance related to shoppable services or price estimator requirements (33%).

That scrutiny led to more than $4 million in civil monetary penalties issued to 14 hospitals that didn’t address their shortcomings, according to the report. Across 2024, the CMS has posted public notice of only one additional civil monetary penalty, for about $871,000.

The CMS told the GAO its enforcement processes would be roughly similar when reviewing for the summer’s new requirements, but plans to use “other approaches in addition to public complaints to identify hospitals for review.” CMS officials also told the office they expect the machine-readable file standardization will make reviews more efficient and potentially open the door for automation.

Still, GAO’s conversations with CMS raised some concerns over whether hospitals are posting usable data. Officials told the watchdog that they do not routinely review hospital files to confirm they are complete nor to check that the posted prices are accurate. They said focusing on “basic reporting requirements … represents the most effective use of agency resources” and that the statute did not require checks of completeness or accuracy.

Based on its investigation, the GAO recommended that “the Administrator of CMS should assess whether hospital price transparency machine-readable files are sufficiently complete and accurate to be usable for supporting CMS’ program goal and implement any additional cost-effective enforcement activities as needed. Such an assessment could include soliciting stakeholder feedback or conducting a study of hospital file completeness and accuracy.”

In a written response, the Department of Health and Human Services said it “concurs” with the recommendation and would “explore the possibility” of such assessments for accuracy and completeness. It would then determine whether to implement any of the “additional cost-effective enforcement activities” floated by the GAO, such as risk-based random sampling.